Re: Using OS disk (SSD) as journal for OSD

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi

Thanks for the reply.
Some follow-up's.

Den 2015-09-12 kl. 17:30, skrev Christian Balzer:
Hello,

On Sat, 12 Sep 2015 17:11:04 +0200 Stefan Eriksson wrote:

Hi,

I'm reading the documentation about creating new OSD's and I'm see:

"The foregoing example assumes a disk dedicated to one Ceph OSD Daemon,
and a path to an SSD journal partition. We recommend storing the journal
on a separate drive to maximize throughput. You may dedicate a single
drive for the journal too (which may be expensive) or place the journal
on the same disk as the OSD (not recommended as it impairs performance).
In the foregoing example we store the journal on a partitioned solid
state drive." From:
http://ceph.com/docs/master/rados/deployment/ceph-deploy-osd/

So I would like to create my journals on the same SSD as I have my OS
(RAID1). Is this good practise to initiate a new disk with:

Which SSDs (model)?
Some SSDs are patently unsuited for OSD journals, while others will have
no issues keeping up with the OS and journal duties.

I'm using SSDSA2BZ10 in my lab, in hardware raid1.


The sda below suggest that your RAID1 is a HW one?
That's a bad choice on two counts, a HW RAID can't be TRIM'ed last I
checked.
And you would get a lot more performance out of a software RAID1 with
journals on both SSDs.
A RAID 1 still might be OK if you can/want trade performance for
redundancy.

I can spare some performance for the convenience to use HW raid on the OS disk.

ceph-deploy disk zap osdserver1:sdb
ceph-deploy osd prepare osdserver1:sdb:/dev/sda

I'm not a ceph-deploy expert or fan, but I'm pretty sure you will need to
create the partitions beforehand and then assign them accordingly.
And using uuids makes things renumbering proof:

ceph-deploy osd prepare ceph-04:sdb:/dev/disk/by-id/wwn-0x55cd2e404b73d348-part4

Christian

It should be ok I think to use just :/dev/sda
(ceph-deploy osd prepare osdserver1:sdb:/dev/sda)

My output is:

[osd03][WARNIN] WARNING:ceph-disk:OSD will not be hot-swappable if journal is not the same device as the osd data [osd03][WARNIN] DEBUG:ceph-disk:Creating journal partition num 4 size 5120 on /dev/sda [osd03][WARNIN] INFO:ceph-disk:Running command: /sbin/sgdisk --new=4:0:+5120M --change-name=4:ceph journal --partition-guid=4:80836633--8e7d-5cf560e53ef --typecode=4:45b09694f-b4b80ceff106 --mbrtogpt -- /dev/sda

So "ceph-deploy osd prepare" seems to recognize the current partitions and create a 4th. Here I ran into bug where the sda did not boot up, but I think its due to --mbrtogpt as the sda was MBR (centos 7 defaults to this when installed on small disks) and not GPT. So the conversion might have gone bad.
So now I will try to force GPT when I install centos7 even on smaller disks.

A question persists about ceph-deploy and the "osd prepare" statement working together with a disk that is already used by the OS. Does it have logic that is sound for this use case? I think it might be a common design choice to use journal and OS on the same disk.
It would be nice to hear some thoughs about the above.

_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com



[Index of Archives]     [Information on CEPH]     [Linux Filesystem Development]     [Ceph Development]     [Ceph Large]     [Ceph Dev]     [Linux USB Development]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]     [xfs]


  Powered by Linux